This may help drive.google.com/file/d/1TSp4...

@lisahuelsmann.bsky.social ky.social @bayceer.bsky.social bsky.social @gfoesoc.bsky.social c.bsky.social @bes-quantitative.bsky.social @nordicoikos.bsky.social @yomoseco.bsky.social @bobohara.bsky.social @florianhartig.bsky.social

Part of the Elite Network of Bavaria, the M.Sc. integrates ecology, statistical modelling, deep learning & AI, remote sensing, and decision science — training graduates to forecast ecological change and communicate it to managers and policy-makers.

The new Master's programme in Ecological Forecasting at UBT and JMU Würzburg is looking to employ six lecturers and a program coordinator!

For more information, please contact Prof. Steven Higgins (www.pfloek.uni-bayreuth.de/en/team/higg...).

I'd love to go, as per usual, but how do people pay for this sort of thing?

$500 for the conference, a flight to Mexico and hotel.. I'm not exactly made of money.

Are there perhaps travel grants for conferences in Germany? Beyond DAAD I haven't really found anything..

Looks like I'm going to @iavs5.bsky.social this summer to tell people about model-based ordination: The One Method That Rules Them All.

#IAVS2026 #gllvm

Calculating and analysing the difference in response of locations, such as the difference in CM_EIVs, creates more problems I won't even touch in this thread.

Community ecology needs to move away from univariate analyses of multispecies data. Multispecies models - GLLVMs, JSDMs- exist, are tractable, facilitate answering questions in a much more statistically appropriate manner.

That doesn't even begin to cover the additional sources of variation introduced by the weights (EIVs)

All of this can only lead to severely biased and erroneous predictions.

Violations including 1) non-normality (for the LMM), 2) improper mean-variance relation (richness has non-constant underdispersion), 3) non-linear relationship with the environment, 4) species' independence, 5) species' homogeneity of niches, 6) lack of definition of the pool (bound for richness).

CM_EIV, like most measures of diversity, is a derived quantity of multispecies data incidence. For species richness alone, a univariate regression doesn't make sense- it requires assumptions that are almost always violated in practice.

This looks like an amazing dataset, exploring an important shift in European plant communities.

Can we please move on from analysing ad-hoc community summaries from multi-species data? The statistical problems run so much deeper than people appreciate.

Excellent visualization of the true distance between the Earth and our moon: #artemis #AGoodPlace

Source: www.reddit.com/r/BeAmazed/s...

No worries, thanks for understanding and not taking this the wrong way. For some reason only citing Niku et al. 2017 has established itself in the literature, despite my repeated attempts for improvement.

I do really enjoy seeing GLLVMs used in practice.

Congrats! Nice to see GLLVMs in practice.

As an aside, please note the citation recommendations under citation("gllvm") or on the github.

The devs put a lot of work in further development and maintenance of the package, and it would be great if that also receives the credit it deserves.

Bar chart titled "Matters of Scale" comparing proposed US research budget cuts to the European Union's €500-million (US$571-million) "Choose Europe" fund. The chart shows: * National Institutes of Health (NIH): $8 billion in cancelled grants and $18 billion in proposed cuts by 2026 (long orange bar). * National Science Foundation (NSF): $5.1 billion in proposed cuts by 2026 (shorter orange bar). * EU's Choose Europe fund: $571 million (very short blue bar). The graphic highlights that the EU fund is much smaller in scale compared to the US budget cuts. Text above the chart explains the EU’s intention to attract US researchers in response to policy decisions by the convicted felon and rapist Donald Trump.

If people think American scientists are somehow going to land in Europe, I've got news for you about the difference between millions and billions.

www.nature.com/articles/d41...

We are looking for a postdoc (up to five years) interested in climate-driven plant extinctions! Working with @manuelsteinbauer.bsky.social and me and a large team on various aspects of Earth system sciences.

More information here: fau-earth-system-science.github.io

I am very excited to contribute to this brand new programme and soon meet the first generation of young ecological forecasters here at #UBT!

@ecoforecast.bsky.social

We are hiring: PhD position in Ecological Data Science We are looking for a PhD student to join the AG Hartig (Theoretical Ecology) in the field of ecological data science / deep learning. 

#PhDPosition in Ecological #DataScience in our group. For details, see www.uni-regensburg.de/universitaet... #MachineLearning #AI #DeepLearning #Statistics #Ecology #AcademicJobs

No worries, this is clear. It confirms what I thought, so my response remains unchanged.

gllvm does not specify the random effects covariance matrix in the same way, it is for the covariate effects instead, so that for a single covariate you get only a variance.

If "trait" is not passed as covariate but as keyword, this may exactly be what gllvm does with formula = ~ (covariate|1), I think as the columns are implicit.

Random effects are assumed to be specific to the columns.

Yes.

What does that mean for the covariance matrix for that random term?

In this article, we present the case for Generalized Linear Latent Variable Models (GLLVMs) as a go-to choice of statistical method for any community ecologist wanting to tackle a range of present-day ecological research questions. GLLVMs bring tools and capabilities from classic (mixed-effects) regression models to multivariate community analysis, providing a number of novel ways to tailor models specifically to one’s study questions and data properties not available when using non-model-based multivariate methods. In order to facilitate further adoption of these methods by community ecologists, we provide 1) a practitioner-focused and practical overview of the advantages the GLLVM framework brings to the table when addressing different core ecological questions, 2) a number of concrete suggestions for how GLLVMs best can be incorporated into the analytical workflow of community ecologists, and 3) two illustrative worked examples of this workflow in action on real-world data.

Nice! A Practitioner’s Guide to Latent Variable Modelling for Community Ecology doi.org/10.32942/X2K... #ecopubs

It may be possible to trick gllvm by smartly formulating the data and adding some NAs, but it's going to be painful.

Unless I misunderstand what MCMCglmm does exactly.

'formula' assumes column-specific random effects with a covariate by covariate covariance matrix.

So here, 'line' is a single covariate, so you'll get a single variance parameter.

No reason why gllvm could not do something else but we did not have the foresight that someone would try this.

The difficulty is that MCMCglmm passes 'trait' as covariate, whereas it also corresponds to the responses?

gllvm is not quite setup to accommodate a structure like that (yet?), as covariates are assumed to not be response-specific.

I'm not very familiar with the MCMCglmm syntax unfortunately, but....

- The 'family' argument is specified in the same way
- an unstructured residual covariance is not possible in gllvm (well, via LVs... so num.lv =..)
- formula = ~(0+line|trait) conceptually corresponds to 'random', I think.

Bluesky works, e-mail works, github discussions works too. I'm not particularly picky.

In case you're looking for an answer from Jenni Niku instead, e-mail or github, because I believe she is not on bluesky.

@rtbecard.bsky.social